研究生: |
蕭宗閔 Tsung-Min Hsiao |
---|---|
論文名稱: |
應用數據增強於卷積神經網路之馬達故障診斷之研究 Applying data augmentation for CNN-based motor faults diagnosis |
指導教授: |
張宏展
Hong-Chan Chang |
口試委員: |
吳瑞南
Ruay-Nan Wu 陳建富 Jiann-Fuh Chen 陳財榮 Tsair-Rong Chen 陳鴻誠 Hung-Cheng Chen |
學位類別: |
碩士 Master |
系所名稱: |
電資學院 - 電機工程系 Department of Electrical Engineering |
論文出版年: | 2020 |
畢業學年度: | 108 |
語文別: | 中文 |
論文頁數: | 84 |
中文關鍵詞: | 故障診斷 、數據增強技術 、卷積神經網路 |
外文關鍵詞: | Fault Diagnosis, Data Augmentation, Convolutional Neural Network |
相關次數: | 點閱:296 下載:0 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
故障診斷在現今工業中扮演重要的角色,若能對設備及早檢測和診斷便可提高系統的穩定度及可靠度,甚至可以避免突發事件的發生以省下維修所帶來的成本損失,因此故障診斷已廣泛應用於各領域,但要實現故障診斷通常需要大量的設備運轉數據來建立一個高準確率且通用於不同任務的故障診斷模型,若訓練數據不足以因應高複雜度的模型則會造成過擬合的現象影響模型的準確率,尤其馬達長期運轉於健康狀態下,使得健康狀態及故障狀態的數據失衡。
本研究致力於應用數據增強技術克服訓練數據不足及不平衡的影響。首先設計四種常見的馬達故障模型,並以一部正常馬達當成健康狀態,量測滿載、半載及無載下此五種馬達狀態之振動加速度訊號並將其轉成三維圖片,接著以在圖像辨識任務中表現良好的卷積神經網路建立故障診斷模型,最後透過兩個案例設計模擬訓練數據不足及不平衡的情形,案例一顯示當訓練數據減少至原本的20%,各負載情形下其準確率皆下降至少40%,案例二顯示當訓練數據不平衡時,各負載情形下其準確率皆未達90%,應用數據增強技術生成圖片將訓練數據增加至足夠及平衡後,兩個案例的準確率皆可提升至90%以上,證明數據增強技術應用於CNN-based馬達故障診斷模型之有效性。
Fault diagnosis plays a major role in modern industry. If faults are detected early in facilities, system stability and reliability can be improved and emergency occurrences can be prevented, thereby decreasing maintenance costs. Fault diagnosis has been practiced in a broad range of disciplines. However, much facility operation data are required to establish a fault diagnosis model that is highly accurate and applicable in different tasks. Insufficient training data sets can result in overfitting for a highly complex model, which decreases model accuracy. Motors, in particular, have long been operated in a normal state, leading to an imbalance between normal and faulty state data.
This study employed data augmentation to mitigate the adverse effects of insufficiency and imbalance in training data. Models on four common types of motor faults were established, and the performance of these modeled faulty motors were compared against that of a normal-functioning motor. The vibration acceleration signals of the five motors at full load, half load, and no load were measured and converted to 3D images. A fault diagnosis model was then constructed using a convolutional neural network, which performs well in image recognition. Finally, two cases were devised to simulate an insufficient and unbalanced training data set, respectively. In Case 1, where the training data set was only 20% the size of the original data set, the fault diagnosis accuracy decreased by at least 40% for each of the aforementioned load conditions; in Case 2, when the training data had an unbalanced number of data points, the fault diagnosis accuracy at each load condition was <90%. When this study’s data augmentation technique was applied to resolve these problems of insufficiency and imbalance, the diagnosis accuracy for both cases increased to >90%. These results confirmed that this study’s data enhancement technology can be effectively employed in a CNN-based model for motor fault diagnosis.
[1] Q. Liu and C. Huang, “A Fault Diagnosis Method Based on Transfer Convolutional Neural Networks,” in IEEE Access, vol. 7, pp. 171423-171430, 2019.
[2] L. Yaguo, Y. Bin, J. Xinwei, L. Naipeng, K. N. Asoke, “Application of machine learning to machine fault diagnosis: A review and roadmap,” Mechanical Systems and Signal Processing, vol. 138, April 2020.
[3] F. Lv, C. Wen, Z. Bao and M. Liu, “Fault diagnosis based on deep learning,” 2016 American Control Conference (ACC), Boston, MA, 2016, pp. 6851-6856.
[4] CNN入門-Overfitting.Available: https://medium.com/@CinnamonAITaiwan/cnn%E5%85%A5%E9%96%80-overfitting-d10acd15ec21
[5] H. Mansourifar, L. Chen and W. Shi, “Virtual Big Data for GAN Based Data Augmentation,” 2019 IEEE International Conference on Big Data (Big Data), Los Angeles, CA, USA, 2019, pp. 1478-1487.
[6] L. Wen, X. Li, L. Gao and Y. Zhang, “A New Convolutional Neural Network-Based Data-Driven Fault Diagnosis Method,” in IEEE Transactions on Industrial Electronics, vol. 65, no. 7, pp. 5990-5998, July 2018.
[7] J. Wang, Z. Mo, H. Zhang and Q. Miao, “A Deep Learning Method for Bearing Fault Diagnosis Based on Time-Frequency Image,” in IEEE Access, vol. 7, pp. 42373-42383, 2019.
[8] Z. Meng, X. Guo, Z. Pan, D. Sun and S. Liu, “Data Segmentation and Augmentation Methods Based on Raw Data Using Deep Neural Networks Approach for Rotating Machinery Fault Diagnosis,” in IEEE Access, vol. 7, pp. 79510-79522, 2019.
[9] A. Fawzi, H. Samulowitz, D. Turaga and P. Frossard, “Adaptive data augmentation for image classification,” 2016 IEEE International Conference on Image Processing (ICIP), Phoenix, AZ, 2016, pp. 3688-3692.
[10] J. Shijie, W. Ping, J. Peiyi and H. Siping, “Research on data augmentation for image classification based on convolution neural networks,” 2017 Chinese Automation Congress (CAC), Jinan, 2017, pp. 4165-4170.
[11] P. Zhang, Y. Du, T. G. Habetler and B. Lu, “A Survey of Condition Monitoring and Protection Methods for Medium-Voltage Induction Motors,” in IEEE Transactions on Industry Applications, vol. 47, no. 1, pp. 34-46, Jan.-Feb. 2011.
[12] H. Tian, Y. Bo-Suk, Y. Zhong-Jun, “Feature-based fault diagnosis system of induction motors using vibration signal,” Journal of Quality in Maintenance Engineering, vol. 13, pp.163-175, June 2007.
[13] W. S. Gongora, H. V. D. Silva, A. Goedtel, W. F. Godoy and S. A. O. da Silva, “Neural approach for bearing fault detection in three phase induction motors,” 2013 9th IEEE International Symposium on Diagnostics for Electric Machines, Power Electronics and Drives (SDEMPED), Valencia, 2013, pp. 566-572.
[14] G.K Singh, S.A. Kazzaz, “Induction machine drive condition monitoring and diagnostic research - A survey,” Electric Power Systems Research, vol. 64, pp. 145-158, 2003.
[15] 徐子權,「電氣與振動檢測法於馬達狀態評估之研究」,碩士論文,台灣科技大學電機工程學系,2014年。
[16] 曾思憲,「感應馬達主動型狀態估測系統之研發」,碩士論文,台灣科技大學電機工程學系,2018年。
[17] 林冠榮,「應用模糊理論與獨立成分分析法於感應電動機之故障診斷系統研製」,碩士論文,台灣科技大學電機工程學系,2017年。
[18] 林芝以,「運用電氣訊號於感應馬達狀態監測與故障診斷之研究」,碩士論文,台灣科技大學電機工程學系,2018年。
[19] C. Zhang, L. Xu, X. Li and H. Wang, “A Method of Fault Diagnosis for Rotary Equipment Based on Deep Learning,” 2018 Prognostics and System Health Management Conference (PHM-Chongqing), Chongqing, 2018, pp. 958-962.
[20] Y. Xu, Y. Sun, J. Wan, X. Liu and Z. Song, “Industrial Big Data for Fault Diagnosis: Taxonomy, Review, and Applications,” in IEEE Access, vol. 5, pp. 17368-17380, 2017.
[21] W. Zhaoxia, L. Fen, Y. Shujuan and W. Bin, “Motor Fault Diagnosis Based on the Vibration Signal Testing and Analysis,” 2009 Third International Symposium on Intelligent Information Technology Application, Shanghai, 2009, pp. 433-436.
[22] G. Lou and H. Shi, “Face image recognition based on convolutional neural network,” in China Communications, vol. 17, no. 2, pp. 117-124, Feb. 2020.
[23] 卷積神經網路(Convolutional neural network, CNN) – CNN運算流程。Available: https://medium.com/@chih.sheng.huang821/%E5%8D%B7%E7%A9%8D%E7%A5%9E%E7%B6%93%E7%B6%B2%E8%B7%AF-convolutional-neural-network-cnn-cnn%E9%81%8B%E7%AE%97%E6%B5%81%E7%A8%8B-ecaec240a631
[24] 卷積神經網路(Convolutional neural network, CNN) – 卷積運算、池化運算。Available: https://medium.com/@chih.sheng.huang821/%E5%8D%B7%E7%A9%8D%E7%A5%9E%E7%B6%93%E7%B6%B2%E8%B7%AF-convolutional-neural-network-cnn-%E5%8D%B7%E7%A9%8D%E9%81%8B%E7%AE%97-%E6%B1%A0%E5%8C%96%E9%81%8B%E7%AE%97-856330c2b703
[25] 卷積神經網路(Convolutional neural network, CNN):卷積計算中的步伐(stride)和填充(padding).Available: https://medium.com/@chih.sheng.huang821/%E5%8D%B7%E7%A9%8D%E7%A5%9E%E7%B6%93%E7%B6%B2%E8%B7%AF-convolutional-neural-network-cnn-%E5%8D%B7%E7%A9%8D%E8%A8%88%E7%AE%97%E4%B8%AD%E7%9A%84%E6%AD%A5%E4%BC%90-stride-%E5%92%8C%E5%A1%AB%E5%85%85-padding-94449e638e82
[26] Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, “Gradient-based learning applied to document recognition,” in Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998.
[27] 使用正則化解決過擬合。Available: https://ljalphabeta.gitbooks.io/python-/content/ch3section4.html
[28] 機器學習中的模型複雜度、過擬合與正則化。Available: https://zhuanlan.zhihu.com/p/68054009
[29] 什麼是人工智慧、機器學習和深度學習? Available: https://medium.com/@chih.sheng.huang821/%E4%BB%80%E9%BA%BC%E6%98%AF%E4%BA%BA%E5%B7%A5%E6%99%BA%E6%85%A7-%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92%E5%92%8C%E6%B7%B1%E5%BA%A6%E5%AD%B8%E7%BF%92-587e6a0dc72a
[30] N. G. Paterakis, E. Mocanu, M. Gibescu, B. Stappers and W. van Alst, “Deep learning versus traditional machine learning methods for aggregated energy demand prediction,” 2017 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT-Europe), Torino, 2017, pp. 1-6.
[31] C. Shorten, T. M. Khoshgoftaar , “A survey on Image Data Augmentation for Deep Learning,” J Big Data, 2019.
[32] Keras Image Data Augmentation各參數詳解。Available: https://www.twblogs.net/a/5b84c5142b71775d1cd1c566